Projections from D2 Neurons in Different Subregions of Nucleus Accumbens Shell to Ventral Pallidum Play Distinct Roles in Reward and Aversion

Abstract

The nucleus accumbens shell (NAcSh) plays an important role in reward and aversion. Traditionally, NAc dopamine receptor 2-expressing (D2) neurons are assumed to function in aversion. However, this has been challenged by recent reports which attribute positive motivational roles to D2 neurons. Using optogenetics and multiple behavioral tasks, we found that activation of D2 neurons in the dorsomedial NAcSh drives preference and increases the motivation for rewards, whereas activation of ventral NAcSh D2 neurons induces aversion. Stimulation of D2 neurons in the ventromedial NAcSh increases movement speed and stimulation of D2 neurons in the ventrolateral NAcSh decreases movement speed. Combining retrograde tracing and in situ hybridization, we demonstrated that glutamatergic and GABAergic neurons in the ventral pallidum receive inputs differentially from the dorsomedial and ventral NAcSh. All together, these findings shed light on the controversy regarding the function of NAcSh D2 neurons, and provide new insights into understanding the heterogeneity of the NAcSh.

Electronic supplementary material

The online version of this article (10.1007/s12264-021-00632-9) contains supplementary material, which is available to authorized users.

Introduction

Many mood disorders are associated with dysfunction of the brain’s reward circuits. Previous studies have identified the striatum as a key player in integrating behavioral responses to both positive and negative reinforcement [1–3]. Approximately 95% of striatum neurons are medium spiny neurons, which are divided into two subpopulations: the dopamine receptor 1-expressing (D1) neurons in the ‘‘direct’’ pathway to the midbrain, and the dopamine receptor 2-expressing (D2) neurons in the ‘‘indirect’’ pathway via the globus pallidus [4–8]. In this canonical model, D1 and D2 neurons function in opposite ways. D1 neurons are traditionally associated with positive reinforcement, approach, and movement initiation, whereas D2 neurons are associated with negative reinforcement, avoidance, and movement termination [3, 9–12]. However, a recent paper reported that both D1 and D2 neurons in the dorsolateral striatum are involved in positive reinforcement, but support different action strategies [13].

The functional role of medium spiny neurons in the nucleus accumbens (NAc) is more complex. NAc D1 neurons project to the ventral pallidum in addition to the midbrain [13]. The NAc has been shown to be important for appetitive motivation for diverse rewards including food, addictive drugs, and brain self-stimulation in animals and humans [9, 14–16]. Some reports also indicate its functions in modulating negative reinforcement, such as punishment, aversion, and avoidance [17–19]. However, selective optogenetic activation of D1 or D2 neurons in the NAc alone is not sufficient to elicit real-time place preference or aversion, although it can promote or suppress cocaine-induced place preference, respectively [9]. Some reports have suggested that NAc D2 neurons play an appetitive and motivational role [15, 20, 21], while another study showed that optogenetic stimulation of NAc D2 neurons generates an ambivalent effect in the self-stimulation task [22]. In addition, a recent study has reported that stimulation of NAc D2 neurons promotes explorative choice behavior [8]. Furthermore, with regard to locomotor activity, basal locomotor activity is reported to be inhibited by photostimulation of NAc D2 neurons [23], while other optogenetics studies have shown that stimulation of NAc D2 neurons has no effect on locomotion [9, 15, 24]. These contradictory results suggest that the functions of NAc D2 neurons are more complicated than previously thought.

Recently, several studies have shown the importance of subregional differences in the NAc shell (NAcSh) [25–28], while most of them focus on D1 neurons. Medial and lateral NAcSh D1 neurons projecting to the ventral tegmental area (VTA) have been identified as parts of different pathways involved in motivated behavior [25]. Dopamine neurons projecting to the medial versus the lateral NAc receive different upstream inputs [28]. Another study showed that D1 neurons in the ventral NAcSh drive aversion, whereas D1 neurons in the dorsal NAcSh drive preference and reward seeking, suggesting that the NAcSh could be divided into dorsal and ventral subzones [27]. Additional evidence for this subdivision comes from a recent study showing that DA terminals in the lateral and ventromedial NAc subdivisions respond to rewarding and aversive stimuli, respectively [26]. All these studies indicate that the differences in NAcSh subregions are important for understanding the functional role of the NAcSh, which might also apply to D2 neurons in the NAcSh. Inspired by these reports, in this study we divided the NAcSh into three subregions, dorsomedial (DMNAcSh), ventromedial (VMNAcSh), and ventrolateral (VLNAcSh), and investigated the functions of D2 neurons separately.

The ventral pallidum (VP) is the primary output of the NAc. NAc D2 neurons exclusively project to the VP, and these projections are inhibitory [5, 29]. The projections from the NAc to the VP have been reported to function in motivation and emotion, such as translating limbic and motivational signals into motor outputs [30, 31], integrating different aspects of addictive behaviors [32], and modulating ‘liking’ and ‘wanting’ [33]. Also, interest in the functional study of VP subpopulations, such as vesicular GABA transporter (VGAT) neurons, vesicular-glutamate transporter 2 (VGLUT2) neurons, and parvalbumin (PV) neurons, has increased recently as they have been shown to be strongly associated with motivation [34, 35], reward and aversion [36, 37], addictive behavior [38], and depression [39]. However, how the functional diversity among NAcSh subregions is related to VP subpopulations is not known.

Here, we hypothesized that D2 neurons in different NAcSh subregions are separately involved in the neural circuits of reward and aversion by innervating different VP subpopulations. Determining the heterogeneity of NAcSh D2 neurons and the functions of their projections to the VP would be an important step towards clarifying the controversy over the function of NAc D2 neurons.

Materials and Methods

Animals and Viruses

Adult (24g–30g) D2-ires-Cre (Stock Number: 017263-UCD, Strain Name: Tg (Drd2 cre) 44 Gsat/M-mcd); D1-ires-Cre (Stock Number: 029178-UCD, Strain Name: Tg(Drd1a-cre)150Gsat/ Mmcd);VGAT-ires-cre (Stock Number: 016962, Strain Name: Slc32a1^tm(Cre)lowl/J); and VGLUT2-ires-cre (Stock Number: 016963, Strain Name: Slc17a6 ^tm(Cre)lowl/J) were used in this study. These transgenic mice have Cre recombinase expression directed to specific neurons. For optogenetic experiments, based on the Cre-loxP recombination system, AAV viruses (AAV1-DIO-ChR2 (H134R)-eYFP, AAV2-EF1a-DIO-GtACR1-P2A-GFP, and AAV1-DIO-eYFP) were injected into a target area, and the functional protein (ChR2 (H134R)-eYFP, GtACR1-P2A-GFP or eYFP) was only expressed in the specific cells containing Cre recombinase [40–42]. In this study, we controlled the location and range of virus expression by the volume, speed, and site of injection of AAV virus. All mice were maintained under the standard conditions of Tsinghua University Animal Facility. All animal procedures were approved by the Animal Care Committee and performed under the Guide for the Care and Use of Laboratory Animals.

Stereotactic Surgery

We conducted all surgery under aseptic conditions using a small animal stereotaxic instrument (RWD Life Science). The mice were anesthetized with 1% sodium pentobarbital (1 g/100 mL). For optogenetic stimulation experiments, AAV viruses (AAV1-DIO-ChR2 (H134R)-eYFP and AAV1-DIO-eYFP) were obtained from the University of Pennsylvania. For optogenetic inhibition experiments, AAV virus (AAV2-EF1a-DIO-GtACR1-P2A-GFP) was obtained from Minmin Luo’s lab. All the viruses were injected at 20 nL/min using a 10-μL Nanofil syringe controlled by the UMP3 and Micro4 system (WPI). We carefully chose anterior NAcSh D2 neurons as our target and divided the NAcSh into DMNAcSh, VMNAcSh, and VLNAcSh zones. Therefore, in D2-ires-Cre mice, the virus (AAV1-DIO-ChR2 (H134R)-eYFP or AAV1-DIO-eYFP) was bilaterally injected to the DMNAcSh (AP: +1.7 mm, ML: ± 0.6 mm, DV: −4.2 mm), VMNAcSh (AP: +1.7 mm, ML: ± 0.6 mm, DV: −4.9 mm), or VLNAcSh (AP: +1.7 mm, ML: ±1.1 mm, DV: −4.9 mm). The optic fibers were bilaterally targeted to the VP (AP: +0.2 mm, ML: ±1.3 mm, DV: −4.8 mm). The virus injection volume was 50 nL. In VGAT-ires-Cre mice or VGLUT2-ires-cre mice, the virus (AAV1-DIO-ChR2 (H134R)-eYFP, AAV1-DIO-eYFP or AAV2-EF1a-DIO-GtACR1-P2A-GFP) was bilaterally injected to the VP (AP: +0.2 mm, ML: ±1.3 mm, DV: −5.1 mm), and the optic fibers were bilaterally targeted to the VP (AP: +0.2 mm, ML: ±1.3 mm, DV: −4.8 mm). The volumes of virus injection were 100 nL for AAV1-DIO-ChR2 (H134R)-eYFP and AAV1-DIO-eYFP, and 200 nL for AAV2-EF1a-DIO-GtACR1-P2A-GFP. After the injection, the needle was left for at least 10 min to make sure the virus spread in the target area before it was slowly withdrawn. The mice remained on an electric blanket until they fully recovered from anesthesia. After surgery, the mice were allowed to recover for at least 3 weeks before initiation of the behavioral protocols.

For rabies-mediated trans-synaptic retrograde tracing experiments [60], AAV viruses (AAV-DIO-EGFP-TVA and AAV-DIO-RG) and rabies virus (SADΔG-dsRed (Enva) were obtained from BrainVTA, Wuhan, China. The viruses were injected using the system described above in a biosafety level 2 (BL2) facility. AAV-DIO-EGFP-TVA and AAV-DIO-RG were mixed in a ratio of 1:2. The mixed viruses were targeted to the DMNAcSh, VMNAcSh or VLNAcSh (all coordinates as above) in D2-ires-cre mice, and the VP in VGAT/VGLUT2-ires-Cre mice. Then rabies virus was injected to the same brain area two weeks later. The volumes were 50 nL for the AAV viruses and 70 nL for the rabies virus in D2-cre mice, and 80 nL for the AAV viruses and 120 nL for the rabies virus in VGAT/VGLUT2-cre mice. The mice were housed in the BL2 facility after surgery, and their brains were harvested for further analysis after one week.

To map monosynaptic inputs to D2 neurons in different NAcSh subregions, we used the Zeiss fully automatic digital slide scanning system (Axion Scan Z1) to collect DsRed, EGFP, and DAPI fluorescent signals from 50-μm frozen sections. To quantify the number of labeled cell in each upstream region, if a cell body in any region was detected in at least 2 consecutive sections, we regarded it as a valid cell. Thus, all the inputs were divided into 120 regions of interest belonging to 11 brain regions, according to a standard brain atlas [43]. The numbers of labeled neurons were manually counted in a double-blind fashion. To calculate the values of input proportion in each brain region, the number of upstream neurons in that region was normalized by the total number of upstream neurons from the whole brain [44, 45, 61].

For the rabies-mediated trans-synaptic retrograde fluorescence in situ hybridization (FISH) experiments, AAV virus (AAV-DIO-mCherry-TVA and AAV-DIO-RG), and rabies virus (SADΔG-EGFP (Enva) were obtained from BrainVTA, Wuhan, China.

Fluorescence In Situ Hybridization

To prepare probes labeled with digoxigenin (DIG)-tagged UTP, the 5’-overhang of the forward primer was modified with the t7 promoter. A linearized cDNA solution of interest was concentrated and purified from PCR production involving mouse whole brain cDNA. Afterwards, DNA production was transcribed by DIG-RNA Labeling Mix (Roche) to generate D2 probes (Accession number NM_010077.1, probe region 49407042 to 49408129) and D1 probes (Accession number NM_010076.1, probe region 54053821 to 54054976).

The fluorescence in situ hybridization (FISH) experiments were developed under the guideline of a previous study [46]. Mice were anaesthetized with 1% sodium pentobarbital and perfused using Diethyl pyrocarbonate (DEPC)-treated phosphate-buffered saline (DEPC-PBS) followed by 4% paraformaldehyde (wt/vol) in DEPC-PBS (DEPC-PFA). Both DEPC-PBS and DEPC-PFA solutions were stored at 4°C until use. Mice were sequentially decapitated, the brains were extracted and post-fixed (DEPC-PFA, 4°C, overnight). After gradient dehydration with 20% and 30% sucrose (wt/vol) in DEPC-PBS overnight, 50-μm coronal sections were cut in Leica bio-systems tissue freezing medium (Leica, UK). Sections were collected in a 6-well cell culture cluster and rinsed with DEPC-PBS. They were further rinsed with DEPC-PTW (DEPC-PBS, 0.1% Tween20), 0.5% Triton, 2× SSC (3 mol/L NaCl, 0.3 mol/L Na₃Citrate·2H₂O), and DEPC-PTW to permeabilize these sections. After acetylation with acetylation buffer and three washes with DEPC-PBS, the sections were transferred to pre-hybridization buffer (50% formamide, 5× SSC, 5 mmol/L EDTA pH 8.0, 0.1% Tween20, 1% CHAPS) for 2 h, and then probes were diluted to 1 μg/mL with hybridization solution (50% formamide, 5 × SSC, 5 mmol/L EDTA pH 8.0, 0.1% Tween20, 1% CHAPS, 300 μg/mL tRNA, 1 × Denhalt’s solution, 1% heparin). The sections were incubated in hybridization solution for 20 h at 65°C, and rinsed with pre-hybridization buffer at 65°C for 30 min after hybridization. Afterwards, the process was repeated with TBST-pre-hybridization buffer (1:1) and TBST. Then the sections were rinsed once in TBST and TAE (1:1), and twice briefly in TAE at room temperature. To further wash off unconjugated probes, the sections were electrophoresed in TAE buffer at 60 V for 2 h, Anti-Digoxigenin-POD (10520200, Roche, 1:500) was applied to the sections, and then they were kept for 30 h at 4°C. After the sections were transferred to TNT, TSA Plus Cyanine 3 (NEL744B001KT, PerkinElmer, 1:100) was applied to detect the primary antibody. Afterwards, the sections were rinsed in DEPC-PBS, and incubated with anti-eYFP (511201, Zen Bioscience, 1:200) at 4°C overnight. For final fluorescence detection, Alexa Fluor 594 GAR (R37117, Life technologies, 2 drops/mL) was applied. Fluorescence was imaged using a confocal microscope (LSM 710 3-channel inverted confocal microscope, Zeiss; LSM 710 META inverted confocal microscope, Zeiss, Germany).

Behavioral Studies

The experimental mice were housed singly on a reverse 12 h/12 h light/dark cycle (lights off at 09:00) at least one week before behavioral experiments and all these experiments were performed during the dark cycle (10:00 to 20:00). Each mouse was only tested in one optogenetic behavioral experiment each day and the experimental sequence was as follows: real time place preference test, open field test, optogenetic two-bottle preference test, progressive-ratio task. We used 10 mW–15 mW, 20 Hz, 5 ms pulses as previously described [25] for optogenetic stimulation experiments. For optogenetic inhibition experiments, we used 1 mW (20 s on /5 s off) in VGAT-Cre mice and 4 mW (20 s on /5 s off) in VGLUT2-Cre mice.

Real Time Place Preference Test

DMNAcSh-D2-ChR2, VMNAcSh-D2-ChR2, VLNAcSh-D2-ChR2, VP-VGAT-ChR2, VP-VGAT-GtACR1, VP-VGLUT2-ChR2, VP-VGLUT2-GtACR1 mice and littermate controls were allowed to run in a custom-made unbiased, balanced two-compartment conditioning apparatus (50 × 25 × 25 cm³) as described previously [47]. One randomly-chosen chamber was paired with photostimulation (473 nm, 10 mW–15 mW, 20 Hz, 5 ms pulses) or photoinhibition depending on the particular experiment(473 nm, 1 mW for VGAT-Cre mice and 4 mW for VGLUT2-Cre mice), and the entry of a mouse triggered photoactivation or photoinhibition. When the mouse entered the other chamber, it did not receive photoactivation or photoinhibition. This behavioral test included two sessions (each lasting 10 min): pre-test (optic OFF) and test (optic ON). Most mice displayed no preference, and those with > 55% side preference on the pre-test (optic OFF) were excluded from further study. The whole experimental process was recorded via a CCD camera interfaced with software coded using MatLab.

Open Field Test

The open field chamber was made of transparent plastic (50 × 50 cm²). Individual mice were placed in the center of the open field and allowed to recover from handling for 5 min before the start of the session. The test lasted 9 min, including three alternating 3-min epochs in chronological order: a pre-test epoch (optic OFF), a test epoch (optic ON) with photostimulation (473 nm, 20 Hz, 10 mW–15 mW, 5 ms pulses) or photoinhibition (473 nm, 1 mW, 20 s on/5 s off in the VGAT-GtACR1 group; and 473 nm, 4 mW, 20 s on/5 s off in the VGLUT2-GtACR1 group), and a post-test epoch (optic OFF). The average speed was calculated by dividing the total movement distance by time in each epoch in MatLab. The whole experimental process was recorded on a CCD camera interfaced with software coded in MatLab.

Progressive Ratio (PR) Test and Fixed Ratio (FR) Training

The PR test was conducted in an operant conditioning chamber (21.5 × 16 × 16 cm³, Anilab Software & Instruments Co., Ltd, China) contained within a sound-attenuating cabinet. The right side of the chamber was fitted with two nose-poke ports, each with an LED in the rear. One port was designated as ‘active’ (triggering delivery of a water reward) and the other as ‘inactive’ (triggering no outcome). Between the “active” and “inactive” ports, there was a port to deliver water rewards. Prior to behavioral sessions, individual mice spent 1 h for contextual habituation. The training procedure and task sequences followed a previous study [15]. Mice had free access to food but had restricted access to drinking water to maintain 80%–85% of their pre-training weight. Before the PR test, the mice were allowed to learn to poke the active port for water rewards in an FR schedule. On the first training session, the mice were trained to nose-poke once for a drop of water for 1 h for a total of seven days until they learned how to acquire the water rewards. In a similar manner, mice were then trained using a FR4 reinforcement schedule for three days until the number of water rewards reached a plateau and then trained on a FR8 schedule for one day. On the test day, the mice were exposed to the PR schedule of reinforcement to obtain water rewards with photostimulation (10–15 mW, 20 Hz, 5 ms pulses, 20 s on/5 s off in D2-Cre-mice) or photoinhibition (4 mW, 20 s on/5 s off in VGLUT2-Cre mice) throughout the session. On the next day, the mice were trained in FR4 to recover to the motivational level for water before the PR post-test. Then the mice executed the post-test session in the PR schedule of reinforcement to obtain water rewards without photostimulation or photoinhibition. The schedule of response criterion (1, 2, 4, 6, 9, 12, 15, 20, 25……) during the PR test was calculated using the following formula (rounded to the nearest integer): [5e^(R*0.2)]−5, where R was equal to the number of water rewards already earned plus 1 (that was the next reinforcer). The breakpoint was defined as the last criterion successfully completed. PR sessions lasted a maximum of 1 h, and failure to make a nose-poke on the active port in any 10-min period resulted in termination of the session. The breakpoint, number of nose-pokes on the active port, and the number of water rewards earned were used as indicators to assess the level of motivation for reward.

Optogenetic Two-Bottle Preference Test

To further assess whether photostimulation/inhibition of the NAcSh and VP caused aversion, we conducted the optogenetic two-bottle preference test adapted from the sucrose preference test as previously described [48, 49]. The chamber (20 × 20 × 22 cm³) was equipped with two contact lickometers connected to two bottles on the box wall. The mouse’s contact of the lickometer was detected via an electronic logic circuit, which relayed digital ‘lick’ signals to a computer via parallel ports. In the test, we assessed whether the aversion caused by the optogenetic activation/inhibition of neurons could counteract the innate preference for ingesting sucrose solution. The mice could make a free choice of licking two contact lickometers to obtain liquid from the two bottles, which each contained water (water bottle) or 2% sucrose solution (sucrose bottle). Mice instinctively preferred sucrose solution to water without paired-laser stimulation/inhibition. The entire test lasted 5 days including two 15-min sessions (10-h interval) per day during the dark phase of mouse activity, and the positions of the two bottles and their associated lickometers were swapped during the second session on each day. The two bottles were filled with water and counter balanced on the first two days, and the basal licking response to the water bottles was measured on day 2. The procedures within the next three days were as follows: Laser OFF (Day 3), Laser ON (Day 4), and Laser OFF (Day 5). The sucrose bottle was coupled to laser stimulation/inhibition upon an animal licking its nozzle. On day 4, the lick signals from the laser-coupled bottle triggered light stimulation (473 nm, 5 ms pulses, 20 Hz for 3 s) or light inhibition (473 nm, constant light for 3 s). The sucrose preference scores were calculated as the ratio of liquid intake, lick number, and lick duration for sucrose solution to the total liquid intake, lick number, and lick duration during each day. If the optogenetic stimulation/inhibition caused aversion, the preference for sucrose solution would shift to water and the ratios would decrease.

Statistics

Statistical analysis was performed using Student’s t tests or one way-ANOVA followed by Sidak multiple comparisons in commercial software (GraphPad Prism; GraphPad Software, Inc., La Jolla, CA). For all results, the significance threshold was set at *P < 0.05; **P < 0.01; ***P < 0.001; ns, non-significant, P > 0.05. All data are shown as the mean ± SEM.

Results

Activation of D2 Neuronal Terminals from NAcSh Subregions to the VP Differentially Induces Preference and Aversion

As the NAcSh has been reported to show functional heterogeneity according to the different coordinates used in previous reports [25–27, 39, 50], we tested whether D2 neurons in different NAcSh regions exert differential functions by differentially innervating the VP. Previous studies have indicated that the anterior and posterior NAcSh show different and often opposing functions [46, 51]. Therefore, we carefully chose anterior NAcSh D2 neurons as our target to avoid the injected virus contaminating the VP and divided the NAcSh into three zones, using the following coordinates as the injection centers for each zone: DMNAcSh AP: +1.7 mm, ML: ±0.6 mm, DV: −4.2 mm, VMNAcSh AP: +1.7 mm, ML: ±0.6 mm, DV: −4.9 mm, and VLNAcSh AP: +1.7 mm, ML: ±1.1 mm, DV: −4.9 mm. To confirm the projections from NAcSh subregions to the VP, we introduced a Cre-inducible viral construct coding the enhanced yellow fluorescent protein (AAV-DIO-eYFP) unilaterally into each of the three subregions in multiple D2-Cre mice (Fig. 1A, C). The results showed that D2 neurons in all three NAcSh subregions from different D2-Cre mice sent projections to a similar anterior area in the VP (Fig. 1B, D). We therefore used this target coordinate in VP for subsequent optogenetic studies.

Fig. 1.

Activation of D2 neuronal terminals from NAcSh subregions to the VP differentially induces preference and aversion. A Schematic of the design of the recombinant AAV-DIO-eYFP virus and injection sites to trace the projection targets of D2 neurons in different subregions of the NAcSh. B Images showing eYFP expression in DMNAcSh (upper left), VMNAcSh (upper middle) and VLNAcSh (upper right) and their projection targets in the VP (lower panels) in D2-Cre mice (scale bars, 200 μm). C, D Images showing eYFP expression in the DMNAcSh (green), VMNAcSh (yellow); and VLNAcSh (red) (C); NAcSh D2 neurons from different subregions all project to the VP with overlapping target areas (D). The green, yellow and red signals are pseudo-colors. Scale bars, 200 μm. The merged images are from multiple mice. E DMNAcSh-D2-ChR2 mice spend more time on the photostimulation-paired side (ChR2: P ON vs OFF < 0.0001; P ChR2 vs eYFP (ON) < 0.0001; n = 8). F VMNAcSh-D2-ChR2 mice spend less time on the photostimulation-paired side (ChR2: ON vs OFF P < 0.0001; ChR2 vs eYFP (ON) P < 0.0001; n = 6). G VLNAcSh-D2-ChR2 mice spend less time on the photostimulation-paired side (ChR2: ON vs OFF P < 0.0001; ChR2 vs eYFP (ON) P < 0.0001; n = 8). H, I Schematic (H) and the time line (I) of the optogenetic two-bottle preference test. J–L Coupling photostimulation of VP-projecting VMNAcSh D2 neurons with sucrose decreases the probability of sucrose intake (J), lick number (K), and lick duration (L) (sucrose intake: P = 0.0002; lick number: P < 0.0001; lick duration: P < 0.0001; n = 6). M–O Coupling photostimulation of VP-projecting VLNAcSh D2 neurons with sucrose decreases the probability of sucrose intake (M), lick number (N), and lick duration (O) (sucrose intake: P < 0.0001; lick number: P < 0.0001; lick duration: P < 0.0001; n = 7). Data represented as the mean ± SEM. All significance values were tested by two-way ANOVA with Sidak’s multiple comparisons test. ***P < 0.001.

Next, to determine the functions of D2 neurons in different NAcSh subregions, we bilaterally injected Cre-inducible adeno-associated virus (AAV) encoding Channel-Rhodpsin2 (AAV-DIO-ChR2 (H134R)-eYFP) into the DMNAcSh, VMNAcSh, and VLNAcSh of D2-Cre mice. We carefully delivered 50 nL virus into the target area to minimize the virus diffusion, which was checked by examining the virus expression after animals were sacrificed (Fig. S1A). As photostimulation of the injection site could activate non-target neurons such as cholinergic interneurons expressing D2 receptors or D2 neurons in the neighboring septal area, we applied axonal photostimulation by implanting an optic fiber over the VP (Fig. S1B). Three weeks after surgery, these animals were subjected to the real time place preference test (RTPP), in which animals’ preference was measured by comparing the proportion of time they spent in the photostimulation-paired side to the total experimental time. The DMNAcSh-D2-ChR2 mice showed preference for the stimulation-paired compartment in comparison with the control group, spending significantly more time in the compartment where they received light stimulation (Fig. 1E). In contrast, activation of VP-projecting VMNAcSh and VLNAcSh D2 neurons induced robust aversion for the stimulation-paired compartment. The proportion of time spent in the photostimulation-paired compartment of the experimental groups was significantly lower than that of control groups (Fig. 1F, G). All control groups spent about equal amounts of time in each compartment (Fig. 1E–G). We also measured the time spent on the stimulation side over time in the RTPP (Fig. S2) to show how the animals responded to the laser simulation and exhibited preference/aversion over the 10 min of the test.

As activation of VMNAcSh or VLNAcSh D2 neuronal terminals in the VP induced movements, we further assessed whether the photoactivation of VP-projecting VMNAcSh or VLNAcSh D2 neurons caused aversion using the optogenetic two-bottle preference test, in which the aversion conferred by optogenetic stimulation shifted the preference away from the innate preference for sucrose [48, 49]. In this test, mice made a free choice of licking two contact lickometers to obtain liquid from bottles containing water or 2% sucrose (Fig. 1H, I). Mice would innately prefer sucrose solution, spending more time consuming sucrose. To determine whether the aversion induced by the activation of VMNAcSh or VLNAcSh D2 neuronal terminals could shift the preference from sucrose to water, laser stimulation (3 s, 10 mW–15 mW, 20 Hz, 5 ms pulses) was coupled to the sucrose lickometer. Behavioral results showed that both test groups significantly shifted their preference from sucrose to water. During the optic ON epoch, the sucrose intake (lick number and lick duration) of VMNAcSh and VLNAcSh ChR2-expressing mice decreased remarkably in comparison with the control groups (Figs. 1J–O and S5A, B).

Optogenetic Activation of VP-Projecting DMNAcSh D2 Neurons Enhances Motivation for Reward

According to recent reports, activation of NAc D2 neurons can lead to increased motivation for reward [15, 20]. Since the activation of VP-projecting DMNAcSh D2 neurons induced a preference for stimulation in the RTPP, we sought to determine whether this projection participates in motivation for reward, using the PR test as described previously [15]. During the training session, mice were trained to learn to poke the active port for water rewards in an FR schedule (Fig. 2A-B), until the number of water rewards earned by each mouse reached a plateau. In the following PR test, laser light was delivered across the whole PR test session (laser-ON PR test). Compared to the control group, the DMNAcSh-D2-ChR2 group showed a significant increase in the number of nose-pokes on the active port and the number of water rewards earned, which translated into a 74.7% increase of the breakpoint (Fig. 2C–E) in the laser-ON PR test. Also, compared to the behavioral result of laser-OFF PR test, mice in the test group showed a significant increase of the breakpoint by 48.54% in the laser-ON PR test (Fig. 2C).

Fig. 2.

Optogenetic activation of VP-projecting DMNAcSh D2 neurons enhances motivation for reward. A, B Schematic of progressive-ratio (PR) test. C DMNAcSh-D2-ChR2 mice show a higher breakpoint during photostimulation in the PR test (ChR2: ON vs OFF P = 0.0005; ChR2 vs eYFP (ON) P < 0.0001, n = 7). D DMNAcSh-D2-ChR2 mice earn more water rewards during photostimulation in the PR test (ChR2: ON vs OFF P = 0.0004; ChR2 vs eYFP (ON) P < 0.0001; n = 7). E DMNAcSh-D2-ChR2 mice show more nose-pokes on the active port during photostimulation in the PR test (ChR2: ON vs OFF P = 0.0002; ChR2 vs eYFP (ON) P < 0.0001; n = 7). Data represented as the mean ± SEM. All significance values were tested by two-way ANOVA with Sidak’s multiple comparisons test. ***P < 0.001.

Altogether, our results suggested that stimulation of VP-projecting D2 neurons in the DMNAcSh increase the motivation of mice for obtaining rewards.

VGAT and VGLUT2 Neurons in the VP Receive Inputs from Different Subregions of the NAcSh

We hypothesized that D2 neurons from different NAcSh subregions innervated different VP subpopulations. By combining trans-synaptic retrograde labeling and FISH, we sought to discern the inputs to VP VGAT and VGLUT2 neurons, which are the two main subpopulations of the VP. Two Cre-inducible AAV helper vectors (AAV-DIO-EGFP-TVA and AAV-DIO-RG) were introduced into the VP of VGAT-Cre and VGLUT2-Cre mice for genetic specificity. Two weeks after AAV transduction, the modified rabies virus, SADΔG-dsRed (Enva), was injected into the starter areas (Fig. 3A). Seven days later, brain sections were cut for imaging. As only the modified rabies virus that co-localized with TVA and RG were able to cross the synapse to label presynaptic neurons, the starter cells could be identified by examining the EGFP signals and co-expression of dsRed (Fig. 3B, D). The neurons immediately presynaptic to the VP were labeled as singly positive for DsRed (Fig. 3C, E).

Fig. 3.

VGAT and VGLUT2 neurons in the VP receive different inputs from different subregions of the NAcSh. A Schematics of the experimental design for cell-type-specific rabies-mediated retrograde tracing. B, C Images from a VGAT-Cre mouse showing AAV-DIO-EGFP-TVA and SADΔG-dsRed (Enva) expression in VP VGAT neurons (B), and SADΔG-dsRed (Enva) labeling of neurons as inputs to VP VGAT neurons in the NAcSh (C). Green, EGFP-TVA; orange, SADΔG-dsRed. Scale bars, 200 μm in B, 500 μm in C. D, E Images from a VGLUT2-Cre mouse showing AAV-DIO-EGFP-TVA and SADΔG-dsRed (Enva) expression in VP VGLUT2 neurons (D), and SADΔG-dsRed (Enva) labeling of neurons as inputs to VP VGLUT2 neurons in the NAcSh (E). Green, EGFP-TVA; orange, SADΔG-dsRed. Scale bars, 200 μm in D, 500 μm in E. F Proportions of neurons as inputs to VP VGAT neurons from distinct subregions within the NAcSh (proportion = number of neurons as inputs/DAPI+ in each zone in the NAcSh. DM vs VM P = 0.0404, DM vs VL P = 0.0035, VM vs VL P = 0.5518; one-way ANOVA; n = 9). G Proportions of neurons as inputs to VP VGLUT2 neurons from distinct subregions within the NAcSh (DM vs VM P = 0.0479, DM vs VL P = 0.0047, VM vs VL P = 0.0321; one-way ANOVA; n = 4). H Ratios of D1 to D2 neurons in NAcSh as inputs to VP VGAT and VGLUT2 neurons (n = 3). I, J Examples of RV virus-labeled neurons as inputs to VP VGAT and VGLUT2, and FISH for D1 and D2 markers in NAcSh. I Green, inputs to VP VGAT neurons; red, FISH for D1 markers in a VGAT-Cre mouse. J Green, inputs to VP VGLUT2 neurons; red, FISH for D2 markers in a VGLUT2-Cre mouse. Blue, green, and red arrows indicate cells that are DAPI, eYFP-positive, and in situ hybridization-positive, respectively; orange arrows show overlapping signals (blue, green, and red). Scale bars, 50 μm. Data represented as the mean ± SEM.

Quantitative analysis revealed that VP VGAT neurons received biased inputs from the ventral NAcSh with sparsely labeled neurons in the dorsomedial NAcSh (Fig. 3C, F). In contrast, afferents to VP VGLUT2 neurons were predominantly located in the dorsomedial NAcSh (Fig. 3E, G). Considering the heavy GABAergic projections from the NAc to the VP, we performed control experiments to rule out non-specific labeling. In the VP, introduction of AAV-DIO-EGFP-TVA and rabies virus without AAV-DIO-RG did not label neurons in the NAc (Fig. S3A, B), indicating that RG was necessary for the retrograde transport of the rabies virus from the starter neurons. Next, to identify the NAcSh neurons labeled by retrograde rabies virus, we performed FISH for D1 or D2 markers in brain sections. In sections from either VGAT-Cre or VGLUT2-Cre mice, rabies-labeled neurons were found to extensively co-localize with D1 or D2 FISH signals across the NAc (Fig. 3I, J), suggesting that VP VGAT and VGLUT2 neurons are innervated by both D1 and D2 neurons. Quantitative analysis further revealed that VP VGLUT2 neurons received inputs from NAcSh D1 and D2 neurons in a ratio of ~1:2, while the ratio of D1 and D2 neuronal inputs to VP VGAT neurons was close to 1:1 (Fig. 3H).

Optogenetic Activation or Inhibition of VGAT and VGLUT2 Neurons in VP Induces Preference and Aversion

We next sought to determine whether VGAT and VGLUT2 neurons in the VP play opposing functional roles in modulating reward and aversion. Besides using AAV-DIO-ChR2 (H134R)-eYFP to realize the activation, we applied Guillardia theta anion channel rhodopsin 1 protein (GtACR1) to inhibit the neuronal activity.

First, we tested the behavioral effect of activating VP VGAT neurons by bilaterally injecting AAV-DIO-ChR2 (H134R)-eYFP into the VP of VGAT-Cre mice and implanting optic fibers over the VP (Figs 4A and S1D). Mice were first tested in RTPP, where VP-VGAT-ChR2 mice exhibited strong preference for the stimulation-paired side, while eYFP-expressing mice spent equal amounts of time on each side of the chamber (Figs 4B and S4A). To explore whether such strong preference could be reversed, we selectively expressed and activated GtACR1 to inhibit VP VGAT neurons. In RTPP, inhibition of VGAT neurons led to a significant reduction of time spent on the laser-paired side (Figs 4C and S4B). GtACR1-expressing mice showed a strong aversive response and preferred to stay in the compartment where they did not receive laser inhibition. Since the inhibition of VP VGAT neurons seemed to lead to a higher level of locomotion, we performed the optogenetic two-bottle preference test to further determine whether inhibition of VP VGAT neurons could robustly induce an aversive response. Laser inhibition (3 s, 1 mW) was coupled to the sucrose lickometer. As predicted, the sucrose intake as well as lick number and lick duration significantly decreased in the VP-VGAT-GtACR1 group, while the preference of eYFP-expressing mice for the sucrose solution did not change (Figs 4D–F and S5C).

Fig. 4.

Optogenetic activation or inhibition of VGAT and VGLUT2 neurons in the VP induces preference and aversion. A Schematic of the design of the recombinant AAV-DIO-ChR2/GtACR1-eYFP virus and injection site to simulate/inhibit VP VGAT neurons in VGAT-cre mice. B VP-VGAT-ChR2 mice spend more time in the photostimulation-paired chamber (ChR2: ON vs OFF P < 0.0001; ChR2 vs eYFP (ON) P < 0.0001; n = 10). C VP-VGAT-GtACR1 mice spend less time in the photoinhibition-paired chamber (GtACR1: ON vs OFF P = 0.0002; ChR2 vs eYFP (ON) P = 0.0001; n = 6). D, F Coupling photoinhibition of VP VGAT neurons to sucrose decreases the probability of sucrose intake (D), lick number (E), and lick duration (F) (sucrose intake: P < 0.0001; lick number: P < 0.0001; lick duration: P < 0.0001; n = 6). G Schematic of the design of the recombinant AAV-DIO–ChR2/GtACR1-eYFP virus and injection site to stimulate/inhibit VP VGLUT2 neurons in VGLUT2-cre mice. H VGLUT2-ChR2 mice spend less time in the photostimulation-paired chamber (ChR2: ON vs OFF P < 0.0001; ChR2 vs eYFP (ON) P < 0.0001; n = 7). I VP-VGLUT2-GtACR1 mice spend more time in the photoinhibition-paired chamber (GtACR1: ON vs OFF P < 0.0001; ChR2 vs eYFP (ON) P < 0.0001; n = 7). J, L Coupling photostimulation of VP VGLUT2 neurons to sucrose decreases the probability of sucrose intake (J), lick number (K), and lick duration (L) (sucrose intake: P < 0.0001; lick number: P < 0.0001; lick duration: P < 0.0001; n = 6). M VGLUT2-GtACR1-VP mice show a higher breakpoint during photoinhibition in the progressive ratio (PR) test (GtACR1: ON vs OFF P < 0.0001; ChR2 vs eYFP (ON) P < 0.0001; n = 7). N VGLUT2-GtACR1-VP mice earn more water rewards during photoinhibition in the PR test (GtACR1: ON vs OFF P < 0.0001; ChR2 vs eYFP (ON) P < 0.0001; n = 7). O VGLUT2-GtACR1-VP mice show more nose-pokes on the active port during photoinhibition in the PR test (GtACR1: ON vs OFF P < 0.0001; ChR2 vs eYFP (ON) P < 0.0001; n = 7). Data represented as the mean ± SEM. All significance values were tested by two-way ANOVA with Sidak’s multiple comparisons test. ***P < 0.001.

We next investigated the function of VP VGLUT2 neurons, making use of the same genetic manipulation (Figs 4G and S1C) as we did in the study of VP VGAT neurons. In RTPP, the VP-VGLUT2-ChR2 group showed a robust avoidance response when entering the stimulation-paired side (Figs 4H and S4C). Next, these ChR2-expressing mice were tested in the optogenetic two-bottle preference task as the activation of VP VGLUT2 neurons seemed to lead to a higher level of locomotion. Laser stimulation (3 s, 10 mW, 20 Hz, 5ms pulses) was coupled to the sucrose lickometer to see whether such an activation decreased the intake of sucrose. In the optic ON epoch, the sucrose intake of ChR2-expressing mice significantly decreased compared to control mice (Figs 4J–L and S5D), suggesting that activation of VP VGLUT2 neurons successfully shifted the preference from sucrose to water. Then we inhibited VP VGLUT2 neurons by expressing and activating GtACR1 to determine whether such inhibition induced or enhanced the positive preference. In RTPP, inhibition of VP VGLUT2 neurons induced clear preference for the laser-paired side (Figs 4I and S4D). Given the behavioral results of manipulation of the DMNAcSh-VP D2 projection, which was suggested to synapse onto VP VGLUT2 neurons according to the tracing results (Fig. 3D, E, G), we tested GtACR1-expressing animals in the PR test to assess the motivation for reward. As expected, consistent with the behavioral performance of DMNAcSh-D2-ChR2 mice, inhibition of VP VGLUT2 neurons led to significant increases in the number of nose-pokes on the active port and the number of water rewards earned, which translated into an 127.87% enhancement of the breakpoint (Fig. 4M–O).

Altogether, our behavioral results indicated that VP VGAT and VGLUT2 neurons drive opposite responses in reward and aversion. In line with the results of activation of VP-projecting D2 neurons in the dorsomedial and ventral NAcSh, the inhibition of VP VGLUT2 neurons drove preference and increased motivation for reward, while the inhibition of VP VGAT neurons was capable of inducing aversion.

Projections from D2 Neurons in Different NAcSh Subregions to the VP Differentially Modulate Movement Speed

In RTPP, activation of VMNAcSh or VLNAcSh D2 neuronal terminals in the VP influenced the locomotion of mice. Hence, we further studied locomotor activity in the open field test. A 9-min test session consisted of three 3-min epochs: pre-test (optic OFF), test (optic ON), and post-test (optic OFF) (Fig. 5A). Interestingly, in comparison with control groups, the VMNAcSh-D2-ChR2 group exhibited significantly higher movement speed during the optic ON epoch than during the OFF epoch (Fig. 5C), whereas the VLNAcSh-D2-ChR2 group showed significantly lower movement speed during the optic ON epoch (Fig. 5D) than during the OFF epoch. Photostimulation of DMNAcSh D2 neuronal terminals in the VP had no effect on movement speed (Fig. 5B).

Fig. 5.

Projections from D2 neurons in different NAcSh subregions to the VP differentially modulate movement speed. A Timeline of the open field test to assess movement speed. B Photostimulation of VP-projecting DMNAcSh D2 neurons has no effect on movement speed during the ON epoch in the open field test (ChR2: ON vs pre-OFF P = 0.9989, ON vs post-OFF P = 0.8554; ChR2 vs eYFP (ON) P = 0.9996; n = 6). C Photostimulation of VP-projecting VMNAcSh D2 neurons increases movement speed during the ON epoch in the open field test (ChR2: ON vs pre-OFF P = 0.0019, ON vs post-OFF P < 0.0001; ChR2 vs eYFP (ON) P < 0.0001; n = 6). D Photostimulation of VP-projecting VLNAcSh D2 neurons decreases movement speed during the ON epoch in the open field test (ChR2: ON vs pre-OFF P = 0.0131, ON vs post-OFF P = 0.0008; ChR2 vs eYFP P = 0.0063; n = 7). E Photostimulation of VGLUT2 neurons in the VP increases movement speed during the ON epoch in the open field test (ChR2: ON vs pre-OFF P < 0.0001, ON vs post-OFF P < 0.0001; ChR2 vs eYFP (ON) P < 0.0001; n = 6). F Photostimulation of VGAT neurons in VP has no effect on movement speed during the ON epoch in the open field test (ChR2: ON vs pre-OFF P > 0.9999, ON vs post-OFF P = 0.9998, n = 6; ChR2 vs eYFP (ON) P = 0.2218; n = 6). G Photoinhibition of VGLUT2 neurons in VP has no effect on movement speed during the ON epoch in the open field test (GtACR1: on vs pre-OFF P = 0. 9894, ON vs post-OFF P > 0.9999; ChR2 vs eYFP(ON) P = 0.9938; n = 7). H Photoinhibition of VGAT neurons in VP shows an increase in movement speed during the ON epoch in the open field test. (GtACR1: ON vs pre-OFF P < 0.0001, ON vs post-OFF P < 0.0001; ChR2 vs eYFP (ON) P < 0.0001; n = 6). Data are presented as the mean ± SEM. All significance values were tested by two-way ANOVA with Sidak’s multiple comparisons test. *P < 0.05; **P < 0.01; ***P < 0.001; ns, P > 0.05, not significant.

Next, we applied the open field assay to determine whether VP VGAT and VGLUT2 neurons are involved in modulating movement speed, since they were differentially innervated by NAcSh subregions. Photoactivation of VP VGLUT2 neurons significantly increased movement speed (Fig. 5E), whereas movement was not influenced by inhibition (Fig. 5G). In contrast, activation of VGAT neurons did not change movement speed (Fig. 5F), while inhibition of VGAT neurons induced a dramatic increase in movement speed (Fig. 5H).

D2 Neurons from Distinct Subregions of the NAcSh Receive Different Inputs from the Whole Brain

Since NAcSh D2 neurons in different subregions have different connection patterns with the VP and different functions, we hypothesized that they may receive distinct inputs as well. To test this, we performed monosynaptic rabies tracing (Fig. 3A). Using the standard brain atlas [43] and a previous report [45] as guidelines to delineate and categorize the brain structures, whole-brain quantitation of dsRed-labeled neurons revealed significant differences in the afferents to D2 neurons in NAcSh subregions.

Regarding the major brain structures, although NAcSh subregions shared similar afferent sources among diverse brain regions (Fig. 6G), quantitative statistics revealed that the proportion of afferent innervation varied in NAcSh subregions. DMNAcSh D2 neurons contained more projections from the septum, hippocampus, and hypothalamus than those in the VMNAcSh and VLNAcSh. The basal forebrain and thalamus both sent predominant projections to the VMNAcSh. Among the olfactory areas, cortex, and amygdala, more neurons were retrogradely labeled from the VLNAcSh than the DMNAcSh or VMNAcSh (Fig. 6A and Table S1).

Fig. 6.

D2 neurons from distinct subregions of the NAcSh receive different inputs from the whole brain. For all the graphs, proportion of total inputs is defined as the percentage of the number of labeled neurons in designated brain areas to the total number of labeled neurons in the whole brain. All abbreviations refer to the standard mouse brain atlas [43]. A Percentages of inputs to D2 neurons in different NAcSh subregions from major brain structures. B Percentages of inputs to D2 neurons in different NAcSh subregions from the anterior cortex. C Percentages of inputs to D2 neurons in different NAcSh subregions from thalamic nuclei. D Percentages of inputs to D2 neurons in different NAcSh subregions from the amygdala and adjacent nuclei. E Percentages of inputs to D2 neurons in different NAcSh subregions from olfactory areas. F Percentages of inputs to D2 neurons in different NAcSh subregions from the hippocampus and posterior cortex. G Representative coronal sections showing labeling of monosynaptic inputs to the DMNAcSh, VMNAcSh, and VLNAcSh. Only the side ipsilateral to the injection site is shown. Scale bar, 1000 μm. Data are presented as the mean ± SEM. All significance values were tested by a mixed model. n = 4 mice for DMNAcSh; n = 5 for VMNAcSh; n = 4 for VLNAcSh, *P < 0.05; **P < 0.01; ***P < 0.001.

Refined patterns of upstream connectivity were revealed by looking into the defined groups of nuclei. In the hippocampus, the DMNAcSh had more inputs from CA1 than the other two subregions (Fig. 6F and Table S2). Also, retrograde signals from the DMNAcSh were dominantly in the dorsal and intermediate parts of the lateral septal nucleus (Fig. S6B and Table S2). Among the thalamic nuclear groups, the anterior paraventricular nucleus contained more upstream neurons for the DMNAcSh and VMNAcSh, while the parafascicular and intermediodorsal (IMD) nuclei contained the fewest upstream neurons for the DMNAcSh. The IMD, centrolateral, paracentral, mediodorsal, and central medial thalamic nuclei had the most upstream neurons for the VMNAcSh (Fig. 6C and Table S2). Also, in the basal forebrain, labeled neurons were predominantly grouped within the VP, which sent out more projections to the VMNAcSh than the DMNAcSh or VLNAcSh (Fig. S6A and Table S2). The VLNAcSh received significantly more inputs in terms of the proportion from specific cortical areas (Fig. 6B, F and Table S2), the piriform cortex (Fig. 6E and Table S2), and specific nuclei of the amygdala, such as the cortex-amygdala transition zone and the anterior cortical amygdaloid nucleus (Fig. 6D and Table S2).

Discussion

In recent years, many reports have revealed the importance of subregional difference within the NAcSh [25–28, 50]. Yang et al. (2018) discovered that activation of lateral NAcSh D1 neurons directly promotes reward-related behaviors, and medial NAcSh D1 neurons are indirectly involved in reward [25]. It was also shown that VTA neurons projecting to the medial and lateral NAcSh have different upstream inputs and produce different behavioral effects [28]. Other recent studies divided the NAcSh into dorsal and ventral zones based on their functions. Dorsal NAcSh dynorphin (co-expressing D1) neurons are involved in the modulation of positive reinforcement, while stimulation of ventral NAcSh dynorphin neurons elicits robust aversive behavior [27]. Another recent study showed that DA terminals in the lateral and ventromedial NAc subzones are excited by rewarding and aversive stimuli, respectively [26]. Inspired by these studies, we hypothesized that NAcSh could be divided into three discrete regions. To test this, we chose NAcSh D2 neurons as our research target, as controversy has been increasing in recent years on whether D2 neurons are responsible for mediating aversion and inhibiting motivation [9, 52] or for increasing motivation [15, 20, 53, 54]. Referring to the coordinates of NAcSh used in the previous literature as well as avoiding contamination of the VP by injected virus, we chose a slightly anterior coordinate at AP +1.7 mm. In most studies of the NAcSh, the coordinates used fall within the range AP +1.4 mm–1.7 mm. Combining optogenetics, trans-synaptic retrograde tracing, and FISH, we identified three subregions of NAcSh D2 neurons based on their functions, connectivity with VP neuron subpopulations, and whole-brain afferent connectivity. Our findings contribute to understanding the diversity of the NAcSh.

Behavioral results showed that D2 neurons in the dorsomedial and ventral NAcSh are involved in the regulation of reward and aversion, respectively. Interestingly, DMNAcSh D2 neurons were specifically involved in reward, as activation of DMNAcSh D2 neuronal terminals in the VP induced preference in RTPP (Figs 1E and S2A) and enhanced motivation for reward during the PR test (Fig. 2C–E). In contrast, activation of VP-projecting VMNAcSh and VLNAcSh D2 neurons both induced robust aversion in RTPP (Figs 1F, G and S2B, C). Behavioral performance in the optogenetic two-bottle preference test showed that photostimulation of VP-projecting VMNAcSh and VLNAcSh D2 neurons both remarkably shifted the preference away from sucrose (Fig. 1J–O), further suggesting a role of ventral NAcSh D2 neurons in regulating aversion. The distinct functional roles of dorsomedial and ventral NAcSh D2 neurons were further confirmed by manipulating their respective downstream targets in the VP (Fig. 3B–G). In line with the result of optogenetically stimulating DMNAcSh D2 neurons, inhibition of VP VGLUT2 neurons induced preference in RTPP (Figs 4I and S4D), and increased the motivation for obtaining reward in the PR test (Fig. 4M–O). Also, in line with the result of stimulating ventral NAcSh D2 neurons, inhibition of VP VGAT neurons induced robust aversion in RTPP and the optogenetic two-bottle preference test (Figs 4C–F and S4B). Furthermore, the activation of VP VGAT neurons could potentially reduce aversion by inhibiting GABAergic neurons in the VTA [36]. While the activation of VP VGLUT2 neurons could potentially constrain reward through projections to the lateral habenula, rostromedial tegmental nucleus, and GABAergic VTA neurons [36, 37]. Thus, our results suggest that the NAcSh can be functionally segregated into dorsomedial and ventral zones based on the distinct functions of D2 neurons in reward and aversion and potentially differential connectivity with VP neuronal subtypes.

NAcSh D2 neurons have traditionally been suggested to mediate the cessation of actions [10, 55, 56], while recent studies suggest much more heterogeneous functions of NAcSh D2 neurons. We suggest that the coordinates of the specific injection sites in these experiments matters. In a report where activation of D2 neurons induces a decrease in locomotor activity [23], the injection site is far more lateral than the sites in other reports (the distance from medial to lateral is >1.1 mm), which is consistent with the behavioral performance of locomotor activity we observed in animals with activation of VLNAcSh D2 neuronal terminals in the VP (AP: +1.7 mm, ML: ±1.1 mm, DV: −4.9 mm) (Fig.5D). Our results suggest that VP-projecting DMNAcSh D2 neurons are not involved in the regulation of movement, and this is supported by recent optogenetic studies in which NAc D2 neurons are not correlated with locomotor activity at a virus injection depth of ~4.3 ± 0.1 mm [9, 15, 24]. A somewhat surprising result was that activation of VMNAcSh D2 neuronal terminals in the VP increased the movement speed (Fig. 5C). However, both VMNAcSh and VLNAcSh D2 neurons are assumed to project inhibitory synapses onto VP VGAT neurons (Fig. 3B, C, 3F), while they differ in the modulation of locomotor activity (Fig. 5C, D). Given that inhibition of VP VGAT neurons by locally expressing GtACR1 significantly increased the movement speed (Fig. 5H), it is possible that VLNAcSh D2 neurons innervate a minor subset of VP GABA neurons whose downstream projection is different from the population innervated by VMNAcSh D2 neurons. Collectively, our data describe differential effects on locomotor activity when ventral NAcSh D2 neurons in different subregions are activated.

Along with the distinct functions of D2 neurons in different NAcSh subregions, our study revealed a new connection pattern from the NAcSh to the VP based on VP cell type, in addition to the canonical model based on topographical organization [34]. Combining retrograde tracing and FISH, we showed that VP VGAT neurons receive more afferent inputs from the ventral NAcSh (Fig. 3B, C, F), while VP VGLUT2 neurons are mainly innervated by the DMNAcSh (Fig. 3D, E, G). Further quantitative analysis revealed that the ratio of NAcSh D1/D2 inputs to VP VGLUT2 neurons is ~1:2 (Fig. 3H). Such a ratio has also been reported in other studies. In an electrophysiological study, 89% of VP neurons have been shown to receive inputs from NAc Core D2 neurons, whereas NAc Core D1 neurons only synapse on 50% of VP neurons [5]. In another report, quantitation of NAc D1 and D2 neuronal inputs to VP PV neurons showed a ratio of 1:2 [39]. However, when we looked into the proportion of D1 and D2 neuronal inputs from the NAcSh to VP VGAT neurons, we found that the D1/D2 ratio was close to 1:1 (Fig. 3H). This evidence suggests that, while the VP proportionally receives more inputs from D2 neurons than D1 neurons, different neuronal subtypes in the VP are differentially targeted by D1 and D2 neurons.

The whole brain input mapping of D2 neurons in NAcSh subregions showed a difference in afferent sources for each subregion (Figs 6A–G and 6S), contributing to understanding the functional divergence of NAcSh D2 neurons. Notably, DMNAcSh D2 neurons received preferential inputs from the ventral hippocampal CA1 region (Fig. 6F, G). As projections from ventral CA1 to the NAc core have been implicated in cocaine conditioned place preference [57], it would be interesting to investigate the specific functional role of CA1 projections to the DMNAcSh. In contrast, ventral NAcSh D2 neurons, shown to function in aversion (Fig. 1F, G, J–O), contained abundant projections from the intermediodorsal thalamic nucleus (Fig. 6C, G), which has recently been shown to be important for processing aversive signals [58]. Hunnicutt and colleagues described comprehensive projection maps for excitatory projections from cortex and thalamus in the striatum [59]. They showed that, for the most focused projection targets, the medial anterior NAc preferentially receives more inputs from the infralimbic cortex and subiculum, while the lateral anterior NAc receives more inputs from the anterior insular and entorhinal cortices, consistent with our results (Fig. 6B, F).

In summary, we investigated D2 neurons in three NAcSh subregions in terms of their inputs/outputs as well as their involvement in reward and aversion. Emphasizing the importance of subregional differences within the NAcSh, this study contributes to sorting out the debates on the function of NAcSh D2 neurons, and provides new insights into understanding the heterogeneity of the NAcSh.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Conflict of interest

All authors claim that there are no conflict of interest.